OpenAI's GPT-Live lets ChatGPT talk and listen at once

OpenAI just shipped GPT-Live, its "smartest voice model yet," rolling out globally across iOS, Android, and ChatGPT.com. The headline trick: a full-duplex architecture that lets ChatGPT listen and speak at the same time—continuously processing your voice while it responds, making dozens of decisions per second about whether to talk, wait, interrupt, or fire a tool.

In practice that means dropped "mhmm" and "got it" acknowledgments, quick back-and-forth, and the ability to just stay quiet while you think. You can cut in mid-sentence or ask it to slow down without breaking the flow. When a question needs web search or deeper reasoning, GPT-Live hands off to GPT-5.5 in the background and folds the answer back into the conversation once it's ready.

Two variants ship: GPT-Live-1 becomes the default for paid Go, Plus, and Pro users, while GPT-Live-1 mini covers the free tier. API access is coming soon.

Why it matters: full-duplex kills the walkie-talkie feel of AI voice—this is the closest ChatGPT has sounded to an actual conversation.